Re: UPDATE on 20 Million Records Transaction or not? - Mailing list pgsql-general

From Ron
Subject Re: UPDATE on 20 Million Records Transaction or not?
Date
Msg-id d9d78de2-21c7-3999-beed-62458347d83b@gmail.com
Whole thread Raw
In response to UPDATE on 20 Million Records Transaction or not?  (Jason Ralph <jralph@affinitysolutions.com>)
Responses RE: UPDATE on 20 Million Records Transaction or not?  (Jason Ralph <jralph@affinitysolutions.com>)
List pgsql-general
On 6/23/20 8:32 AM, Jason Ralph wrote:

Hello List,

PostgreSQL 11.8 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.4.7 20120313 (R

ed Hat 4.4.7-23), 64-bit

 

I am planning an update on a table with 20Million records, I have been researching the best practices.  I will remove all indexes and foreign keys prior to the update, however I am not sure if I should use a transaction or not.

My thought process is that a transaction would be easier to recover if something fails, however it would take more time to write to the WAL log in a transaction. 


Are you updating every row in the table?

Are you updating indexed fields?  (If not, then leave the indexes and FKs, since they won't be touched.)

 

Would it make sense to make a back up of the table then execute update without a transaction?


Always make a backup.

How would you guys do it?


It depends on what percentage of the rows are being updated, which columns are being updated and how big the records are.


--
Angular momentum makes the world go 'round.

pgsql-general by date:

Previous
From: Adrian Klaver
Date:
Subject: Re: pg_dump empty tables
Next
From: Jason Ralph
Date:
Subject: RE: UPDATE on 20 Million Records Transaction or not?